events-extraction

name	events-extraction
description	Extracts and formats upcoming events and webinars from provided URLs. Use when running Phase 2 of the newsletter pipeline after the Phase 1C discoveries receipt exists. Fetches event pages, classifies into canonical categories, formats virtual and in-person event tables. Keywords: events extraction, phase 2, webinars, conferences, event formatting.
metadata	{"category":"domain","phase":"2"}

Events Extraction

Extract, classify, and format upcoming events and webinars for the newsletter.

Quick Start

1. Receive list of event URLs (5-15 typical)
2. Run python3 tools/extract_event_sources.py <START_DATE> <END_DATE>
3. Record the phase2_event_sources receipt for the generated JSON
4. Fetch each URL and extract event details
5. Classify: Virtual, In-person, or Hybrid
6. Assign canonical categories (max 2, prefer 1)
7. Format into tables and bullet lists
8. Write output to workspace/newsletter_phase2_events_YYYY-MM-DD.md

Sequencing: This phase starts only after Phase 1C discoveries have been written and receipted.

Inputs

• Deterministic Phase 2 event-source artifact (mandatory):
  • workspace/newsletter_phase2_event_sources_YYYY-MM-DD.json
  • Generated by: python3 tools/extract_event_sources.py <START_DATE> <END_DATE>
• Deterministic source config (required by extractor):
  • kb/EVENT_SOURCES.yaml
• Optional direct URLs (fallback only when artifact is missing):
  • https://github.com/resources/events
  • https://developer.microsoft.com/en-us/reactor/series/S-1625/
  • https://developer.microsoft.com/en-us/reactor/series/S-1631/
  • Note: Reactor series IDs are curated seeds and should be refreshed during KB maintenance.
• Curator notes file (optional): workspace/Jan.md or similar
• COLLECTION_DATE: Date for filename (defaults to today)

Output

• File: workspace/newsletter_phase2_events_YYYY-MM-DD.md

Core Workflow

Step 0: Deterministic Source Extraction

Run deterministic extraction before event curation:

python3 tools/extract_event_sources.py <START_DATE> <END_DATE>

Then verify output exists:

• workspace/newsletter_phase2_event_sources_<END_DATE>.json

If this phase is part of a full run, record receipt immediately after writing the JSON:

bash tools/record_phase_receipt.sh <START_DATE> <END_DATE> phase2_event_sources workspace/newsletter_phase2_event_sources_<END_DATE>.json

Step 1: Discovery

Use the deterministic JSON candidates as primary inputs. Scan all candidate URLs, then enrich with in-range event detail pages.

GitHub Resources Events (github.com/resources/events):

• Fetch the page and event deep links (github.com/resources/events/<slug>), extract all upcoming events
• These are primarily in-person events (AI Tour, GitHub Connect, conferences) and KUWC virtual series

Microsoft Reactor (series pages + event deep links):

• Prefer series pages and deep links from the JSON artifact
• Follow series links (for example S-1625 Agentic DevOps Live) to gather session pages
• Avoid Reactor search URLs as row-level event links

Microsoft AI Tour (aitour.microsoft.com):

• Check for cities where GitHub has a confirmed booth/presence
• Cross-reference with github.com/resources/events (AI Tour events appear on both)

Curator notes: Check for manually added event URLs or KUWC series entries

Step 1.5: Reactor Event Filtering (L69)

The Reactor catalog contains many events. Apply these filters:

INCLUDE (all must be true):

• Event is in English
• Event involves GitHub, Copilot, GHAS, VS Code, or Azure DevOps with GitHub integration
• Event is relevant to enterprise engineering leaders (not consumer/hobbyist)
• Event provides actionable training, product updates, or security guidance

INCLUDE event types:

• Agentic DevOps Live series (GitHub + Azure integration, modernization, security)
• VS Code Live release events (product updates)
• AI Dev Days with GitHub/Azure tracks (hackathons with enterprise prizes)
• Security sessions using GHAS + Defender for Cloud
• App modernization with Copilot agents (Java, .NET)
• GitHub + Azure DevOps workflow sessions

EXCLUDE:

• Non-English language events (Spanish, Portuguese, etc.)
• Competition/gaming formats ("Agents League Battle", esports-inspired)
• Generic AI framework tutorials without GitHub integration (LangChain4j, generic Python AI)
• Beginner-level series on non-GitHub tools
• Events with no GitHub speaker or GitHub product involvement

Key signal: Check the event tags. Events tagged with Github, GitHub Copilot, Agentic DevOps, or VS Code are strong candidates. Events tagged only with AI or generic topics without GitHub involvement should be excluded.

Step 2: Extraction

Extract required fields per event:

• Event title (exact from source)
• Date (normalize to date only for virtual; include CT times only for conference sessions)
• Format: Virtual, In-person, or Hybrid
• Registration URL (direct sign-up link)

Optional but recommended:

• Location (City, ST, Country for in-person)
• Description (1-2 sentences, enterprise value focused)
• Speakers (if available)
• Series information

Step 3: Classification

See events-formatting.md for complete formatting rules.

Canonical categories (use ONLY these, prefer 1, max 2):

• Copilot: All GitHub Copilot related sessions (VS Code AI events implicitly map here)
• GitHub Platform: Core platform, Actions, general feature launches
• Developer Experience: Workflows, training, productivity enablement
• Enterprise: Compliance, governance, administration, security policy

Skill level: Include only "Introductory" or "Advanced" when explicitly stated. Omit otherwise.

Step 4: Formatting

Virtual events: ## Virtual Events heading, table with Date + Event + Categories (NO times) In-person events: ## In-Person Events heading, table with Date + Location + Event Behind the scenes: ## Behind the scenes heading, short bullets for operational notes/recordings Hybrid events: List under In-person with "(Hybrid, virtual attendance available)" Conference sessions: Detailed tables with Date + Time (CT) + Session + Description

Step 5: Standard Content

Always include these blocks in the output:

• Copilot Fridays back catalog link
• Brian's YouTube playlists (3 playlists)

Step 6: Validation

Before writing output:

• Title, date, and registration URL present for all events
• Conference session times converted to CT
• All links embedded as markdown (no raw URLs)
• Hybrid listed under In-person with virtual note
• Categories use canonical taxonomy: Copilot, Copilot, Advanced, Copilot, New Features, GHAS, GHAS, Security, Enterprise, GitHub Platform. Never use Developer Productivity or Platform alone.
• Enterprise relevance satisfied
• Source diversity present: include event links from both GitHub Resources and Microsoft Reactor when available in range
• At least one Reactor-linked event appears in Virtual Events for >=30-day ranges when Reactor has in-range GitHub events
• Event count floor by range length:

  • =60-day range: 12+ total events

  • =30-day range: 8+ total events

  • <30-day range: 4+ total events
• Event row URLs are deep links, not landing/search placeholders
• Do not use these as row-level event URLs:
  • https://resources.github.com/events/
  • https://github.com/resources/events
  • https://developer.microsoft.com/en-us/reactor/?search=...
  • https://resources.github.com/copilot-fridays-english-on-demand/

Reference

• Events Formatting - Table specs, category mapping, timezone rules
• Benchmark Example - Known-good Dec 2025 events

Done When

• Event-source artifact exists at workspace/newsletter_phase2_event_sources_*.json
• Events file exists at workspace/newsletter_phase2_events_*.md
• All provided URLs processed
• Virtual events in table format (date only, no times)
• In-person events in table format (Date, Location, Event)
• ## Behind the scenes section present with short notes
• Canonical categories only (Copilot, GitHub Platform, Developer Experience, Enterprise)
• Standard content blocks included (Copilot Fridays, YouTube playlists)
• Event count floor met for date range
• At least one Reactor-sourced event included when Reactor has in-range enterprise events
• No raw URLs in output